DataAnalytics-Project

This repo is for the Semester 5 Project for the course UE18CS312: Data Analytics

ClassicXI – Optimal IPL Team Selection based on Player Rating and Venue Based Analysis

The final report for this project can be found here
The plagiarism check report (6%) for this project can be found here
The stage1 report can be found here
Datasets used can be found here here

Abstract

Indian Premier League (IPL) is one of the most prestigious T20 cricket tournaments across the world. This paper focuses on suggesting the playing XI squad given a squad of 15 players. We propose a model that rates all the players in the IPL and based on their predicted rating suggests the best playing XI combination the team can select for a given match at a given venue. Our model experiments with various distance similarity measures and other statistics to provide the best possible playing XI. It also provides insights on toss decision at a given venue based on history.

Setup

Dependencies

Make sure you have numpy, pandas, pandasql, json, sklearn and matplotlib of the latest versions installed.
If not, pip install <library_name> or conda install <library_name> after conda update conda

Running the program

1. Copy the DA_Project.ipynb and all the 3 datasets into a folder and then,
   cd into that folder
   Run jupyter notebook and open the file DA_Project.ipynb

2. Run all cells. It should take around 5 minutes to finish running all the cells.

Rating Algorithms

I. Modelling Batsmen

Input: Players data from IPL 2008 to 2019 seasons, IPL Career Stats
Output: Batsman Rating

Features for the Batsman model:

1. Number of innings
2. Batsman Average
3. Strike Rate
4. Fifties
5. Hundreds
6. Number of 4s and 6s

Algorithm

for all players in the dataset:
    
    // All time Form ( IPL 2008 to 2019 )
    matches = Number of innings batted in (Min Max Normalization)
    achievement_weight = (20 x Hundreds) + (10 x Fifties) + (3 x sixes) + Fours
    career_stat_score = (0.3 x achievement_weight) + (0.55 x Batsman Average) + (0.15 x Strike Rate)
    Overall_Career_Score = (matches x career_stat_score)

end for


for all players in the dataset:
    
    // Recent Form ( IPL 2018 to 2019 )
    matches = Number of innings batted in (Min Max Normalization)
    achievement_weight = (20 x Hundreds) + (10 x Fifties) + (3 x sixes) + Fours
    stat_score = (0.3 x achievement_weight) + (0.55 x Batsman Average) + (0.15 x Strike Rate)
    Recent_Score = (matches x stat_score)

end for

for all players in the dataset:
    Overall_Career_Score = Overall_Career_Score / max(Overall_Career_Score)
    Recent_Score = Recent_Score / max(Recent_Score)
    Batsman Rating = (0.4 x Overall_Career_Score) + (0.6 x Recent_Score)
end for

Min Max Normalise( Batsman Rating ) X 100

II. Modelling Bowler

Input: Players data from IPL 2008 to 2019 seasons, IPL Career Stats
Output: Bowler Rating

Features for the Bowler model:

1. Matches - Number of Innings
2. Bowling Economy
3. Bowling Average
4. 4 wicket Hauls
5. 5 wicket Hauls
6. Wickets Taken
7. Maiden overs
8. Number of dot balls

Algorithm

for all players in the dataset:
    
    // All time form ( IPL 2008 to 2019 )
    matches = Number of innings bowled in (Min Max Normalization)
    wicket_weight = (30 x 5_wicket_haul) + (20 x 4_wicket_haul) + (10 x Wickets_Taken)
    achievement_weight = (5 x Maiden_Overs) + dot_balls
    career_stat_score = (Bowling Average) + (Bowling Economy) + (1 / 1 + achievement_weight)
    Overall_Career_Score = (matches x wicket_weight) / (career_stat_score) 

end for

for all players in the recent form dataset:
   
    // Recent Form ( IPL 2018 and 2019 ) 
    matches = Number of innings bowled in (Min Max Normalization)
    wicket_weight = (30 x 5_wicket_haul) + (20 x 4_wicket_haul) + (10 x Wickets_Taken)
    achievement_weight = (5 x Maiden_Overs) + dot_balls
    stat_score = (Bowling Average) + (Bowling Economy) + (1 / 1 + achievement_weight)
    Recent_Score = (matches x wicket_weight) / (stat_score) 
    
end for

for all players in the dataset:
    Overall_Career_Score = Overall_Career_Score / max(Overall_Career_Score)
    Recent_Score = Recent_Score / max(Recent_Score)
    Bowler Rating = (0.35 x Overall_Career_Score) + (0.65 x Recent_Score)
end for

Min Max Normalise( Bowler Rating ) x 100

Team Members:

1. Lavitra Kshitij Madan
2. Ritik Hariani
3. Arjun Chengappa
4. Aditya G Burli